Our invention relates to the conversion of electric signals from digital to analog form. More specifically, our invention pertains to a method of, and a system for, translating a digital data signal such as a digitized audio or video signal into an equivalent analog signal, with the addition of digital dither (an artifically created white noise signal) to the digital data signal and the subsequent removal of the dither from the analog data signal, with a view to the reduction of noise and distortion. Still more specifically, our invention concerns improvements in such a method and system whereby the digital dither to be added to the data signal is suppressed or reduced in level, as required, for the digital to analog coversion of the data signal with a greater dynamic range than heretofore under the same limitations of the hardware in use.
The pulse code modulation (PCM) or digital processing of audio signals has become, or is becoming, the mainstream of high fidelity sound recording and reproduction with the advent and ever increasing commercial acceptance of compact discs (CDs). However, some problems remain unsolved, or not perfectly solved, in the art of digital sound processing. One of these is the "quantization noise", that is, the differences between the samples of the music wave and the quantized values of the samples. The quantization noise becomes particularly pronounced, distorting the reproduced sound as higher harmonics, when the input signal level is low and there are a relatively small number of quantization steps. Even when the input signal level is high, the quantization noise will distort the signal if it changes slowly.
We are aware of a conventional solution to this quantization noise problem. The solution involves the use of dither for turning the quantization noise into white noise which hardly affects the appreciation of the reproduced sound. A digital dither signal is superposed on the audio signal prior to its digital to analog conversion and, following the conversion, is removed from the analog audio signal. This technique is discussed extensively in the article entitled "The Application of Large Amplitude Dither to the Quantization of Wide Range Audio Signals" by Yoshio Yamasaki in The Journal of the Acoustical Society of Japan, Vol. 39, No. 7, published 1983.
We have found that this known solution has a weakness. As digital dither is added to the digital audio or other data signal, the total level of the resulting data and dither signal may exceed the capacity (maximum allowable input level) of the digital to analog (D/A) converter used for the conversion of the data and dither signal, particularly when the data signal level is high. There might be contemplated the use of a D/A converter, as well as an adder for the addition of data and dither, of sufficiently large capacity to accept the expected maximum level (number of bits) of the combined data and dither signal. Such hardware is very expensive and so impractical.
It is more practical to employ a D/A converter and adder of minimal capacities required; for example, the converter and adder may be each constructed to accept only the number of bits (e.g. sixteen) of which the digital data signal is composed. The use of such limited capacity converter and adder has conventionally necessitated the reduction, as required, of the level of the data signal when it is combined with the digital dither by the adder, to such an extent that the combined data and dither signal may not overflow the adder. This of course incurs the sacrifice of the dynamic range to the detriment of the high fidelity reproduction of music.